Skip to content

trace-summary: record message bytes per entry method per interval #3937

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
lvkale wants to merge 1 commit into
base: main
Choose a base branch
from
Open

trace-summary: record message bytes per entry method per interval #3937

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
46 changes: 43 additions & 3 deletions src/ck-perf/trace-summary.C
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -218,6 +218,7 @@ SumLogPool::~SumLogPool()
delete[] epInfo;
delete[] cpuTime;
delete[] numExecutions;
delete[] msgBytes;
}

void SumLogPool::addEventType(int eventType, double time)
Expand Down Expand Up @@ -272,13 +273,17 @@ void SumLogPool::initMem()

cpuTime = NULL;
numExecutions = NULL;
msgBytes = NULL;
if (sumDetail) {
cpuTime = new double[poolSize*epInfoSize];
_MEMCHECK(cpuTime);
memset(cpuTime, 0, poolSize*epInfoSize*sizeof(double));
numExecutions = new int[poolSize*epInfoSize];
_MEMCHECK(numExecutions);
memset(numExecutions, 0, poolSize*epInfoSize*sizeof(int));
msgBytes = new CmiUInt8[poolSize*epInfoSize];
_MEMCHECK(msgBytes);
memset(msgBytes, 0, poolSize*epInfoSize*sizeof(CmiUInt8));

// int i, e;
// for(i=0; i<poolSize; i++) {
Expand Down Expand Up @@ -453,6 +458,30 @@ void SumLogPool::write(void)
}
if (count > 1) fprintf(sdfp, "+%d", count);
fprintf(sdfp, "\n");
// Write out msgBytes
// Run length encoding (RLE) along EP axis
fprintf(sdfp, "MsgBytesPerEPperInterval ");
CmiUInt8 lastBytes = getMsgBytes(0,0);
count=0;
fprintf(sdfp, "%llu", (unsigned long long)lastBytes);
for(e=0; e<_numEntries; e++) {
for(i=0; i<numBins; i++) {

CmiUInt8 u = getMsgBytes(i, e);
if (lastBytes == u) {
count++;
} else {

if (count > 1) fprintf(sdfp, "+%d", count);
fprintf(sdfp, " %llu", (unsigned long long)u);
lastBytes = u;
count = 1;
}
}
}
if (count > 1) fprintf(sdfp, "+%d", count);
fprintf(sdfp, "\n");

// Write out numExecutions
// Run length encoding (RLE) along EP axis
fprintf(sdfp, "EPCallTimePerInterval ");
Expand Down Expand Up @@ -525,7 +554,8 @@ void SumLogPool::setEp(int epidx, double time)

// Called once from endExecute, endPack, etc. this function updates
// the sumDetail intervals.
void SumLogPool::updateSummaryDetail(int epIdx, double startTime, double endTime)
void SumLogPool::updateSummaryDetail(int epIdx, double startTime, double endTime,
CmiUInt8 bytes)
{
if (epIdx >= epInfoSize) {
CmiAbort("Too many entry points!!\n");
Expand Down Expand Up @@ -557,7 +587,10 @@ void SumLogPool::updateSummaryDetail(int epIdx, double startTime, double endTime
CmiAbort("Error: end time of EP is less than start time\n");
}

// startingBinIdx has been walked forward to the bin the execution ended
// in; count the message, and its size, where the run is counted.
incNumExecutions(startingBinIdx, epIdx);
addToMsgBytes(startingBinIdx, epIdx, bytes);
}

// Shrinks pool[], cpuTime[], and numExecutions[]
Expand All @@ -576,12 +609,14 @@ void SumLogPool::shrink(void)
for (int e=0; e < epInfoSize; e++) {
setCPUtime(i, e, getCPUtime(i*2, e) + getCPUtime(i*2+1, e));
setNumExecutions(i, e, getNumExecutions(i*2, e) + getNumExecutions(i*2+1, e));
setMsgBytes(i, e, getMsgBytes(i*2, e) + getMsgBytes(i*2+1, e));
}
}
// zero out the remaining intervals
if (sumDetail) {
memset(&cpuTime[entries*epInfoSize], 0, (numBins-entries)*epInfoSize*sizeof(double));
memset(&numExecutions[entries*epInfoSize], 0, (numBins-entries)*epInfoSize*sizeof(int));
memset(&msgBytes[entries*epInfoSize], 0, (numBins-entries)*epInfoSize*sizeof(CmiUInt8));
}
numBins = entries;
CkpvAccess(binSize) *= 2;
Expand Down Expand Up @@ -647,6 +682,7 @@ TraceSummary::TraceSummary(char **argv):msgNum(0),binStart(0.0),idleStart(0.0),
_logPool = new SumLogPool(CkpvAccess(traceRoot));
// assume invalid entry point on start
execEp=INVALIDEP;
execMsgBytes = 0;
inIdle = 0;
inExec = 0;
depth = 0;
Expand Down Expand Up @@ -685,7 +721,9 @@ void TraceSummary::beginExecute(envelope *e, void *obj)
beginExecute(-1,-1,_threadEP,-1);
}
else {
beginExecute(-1,-1,e->getEpIdx(),-1);
// the message length is passed on so that summary detail can record how
// many bytes an entry method received, not only how many messages
beginExecute(-1,-1,e->getEpIdx(),-1,e->getTotalsize());
}
}

Expand Down Expand Up @@ -725,6 +763,7 @@ void TraceSummary::beginExecute(int event,int msgType,int ep,int srcPe, int mlen
*/

execEp=ep;
execMsgBytes=(mlen > 0) ? (CmiUInt8)mlen : 0;
double t = TraceTimer();
//CmiPrintf("start: %f \n", start);

Expand Down Expand Up @@ -796,9 +835,10 @@ void TraceSummary::endExecute()
binTime += t - ts;

if (sumDetail && execEp >= 0 )
_logPool->updateSummaryDetail(execEp, start, t);
_logPool->updateSummaryDetail(execEp, start, t, execMsgBytes);

execEp = INVALIDEP;
execMsgBytes = 0;
}

void TraceSummary::endExecute(char *msg){
Expand Down
17 changes: 16 additions & 1 deletion src/ck-perf/trace-summary.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -184,6 +184,12 @@ class SumLogPool {
/// for Summary-Detail
double *cpuTime; //[MAX_INTERVALS * MAX_ENTRIES];
int *numExecutions; //[MAX_INTERVALS * MAX_ENTRIES];
/** Bytes of the messages counted by numExecutions, so that a summary
trace can say how much data an entry method received and not only how
many messages. Only messages delivered to an entry method are counted;
packing and unpacking, which numExecutions also counts, contribute
none. */
CmiUInt8 *msgBytes; //[MAX_INTERVALS * MAX_ENTRIES];

public:
SumLogPool(char *pgm);
Expand Down Expand Up @@ -232,10 +238,18 @@ class SumLogPool {
inline int incNumExecutions(unsigned int interval, unsigned int ep){
++numExecutions[interval*epInfoSize+ep];
return numExecutions[interval*epInfoSize+ep]; }
inline CmiUInt8 getMsgBytes(unsigned int interval, unsigned int ep){
return msgBytes[interval*epInfoSize+ep]; }
inline void setMsgBytes(unsigned int interval, unsigned int ep, CmiUInt8 val){
msgBytes[interval*epInfoSize+ep] = val; }
inline CmiUInt8 addToMsgBytes(unsigned int interval, unsigned int ep, CmiUInt8 val){
msgBytes[interval*epInfoSize+ep] += val;
return msgBytes[interval*epInfoSize+ep]; }
inline int getUtilization(int interval, int ep);


void updateSummaryDetail(int epIdx, double startTime, double endTime);
void updateSummaryDetail(int epIdx, double startTime, double endTime,
CmiUInt8 bytes = 0);


};
Expand All @@ -250,6 +264,7 @@ class TraceSummary : public Trace {
int execEvent;
int execEp;
int execPe;
CmiUInt8 execMsgBytes; /* size of the message being executed, for summary detail */
int msgNum; /* used to handle multiple endComputation calls?? */

/* per-log metadata maintained to derive cross-event information */
Expand Down
Loading